Drill Down in hierarchy inside a matrix - performance depending on relationship

[hai_pbi]

Hi,

I share here a very simple mock example to contextualize my question. 
I have two table, MainTable having in Value a serie from 1 to 10'000
And DetailTable having a thiner granularity -> Two IDs for each ID of Maintable 

I create a Matrix for which the main granularity at the row level is on MainTable IDs (Value column), but I also want to drill down back and forth inside the detail table, as depicted here :


Question : 
* If the relationship between the table is uni-directional, the time to drill down in the hierarchy is dependent on the numbers of rows I show on the matrix. 

* If I put the relationship to bi-directional, the processing time it's not dependent on the size of the tables and is very rapid

Why ? What is the logic underneath ? Why with a unidirectional relationship I can still retrieve the TestField (at the MainTable level) ?

Thank you in advance for any inputs 🙂

Mock pbi file : https://we.tl/t-QgPOZ0EAse

[lbendlin]

"If the relationship between the table is uni-directional, the time to drill down in the hierarchy is dependent on the numbers of rows I show on the matrix."

That's not how rendering works. The number of rows shown in the visual most of the time does not exceed 501.  You can see the actual query in the performance amalyzer.

You may want to read the white papers on Vertipaq, the Storage Engine/Formula engine concept, query plans etc.